I very narrowly mean “avoiding python” as an interoperability layer if it isn’t strictly necessary. Python itself has many great uses, and I spend half my time using it. But at any point in time, there is a 50/50 chance that my conda setup is broken or outdated. When I am working on a python project it is annoying but fine because I can setup virtual environments, etc. to make things reproducible. Then my CI instantiates that environment/requirements.txt, etc. and it is pretty reproducible. But anyone who has done this with python called from Julia knows how fragile it is to have both working on machines, continuous integration servers, etc. This is all python setup and environment fragility and has nothing to do with Julia.
All of that is worthwhile if there is a package which is inherently connected to python which makes that burden worthwhile. But this package is for splines, and there are already lots of packages in Julia which have splines (even if they aren’t as good as yours). If yours is better, then I would love to try it out, but not if it means adding in a whole python environment and its ensuring maintenance/reproducibility issues just to do so.
Anyways, that is my last comment on this. I would absolutely love to have another spline/interpolation package in the julia ecosystem, and I really hope your wrapper is successful.